AR1seg-package: Segmentation of an AR(1) Gaussian process

In AR1seg: Segmentation of an autoregressive Gaussian process of order 1

Description

This package consists in an implementation of a robust approach to solve the problem of multiple change-point estimation in the mean of a Gaussian AR(1) process. A robust estimator of the autoregression parameter is proposed and used to build a decorrelated series on which a classical penalized least-square approach is applied.

Details

Package:	AR1seg
Type:	Package
Version:	1.0
Date:	2014-06-04
License:	GPL-2

Author(s)

S. Chakar, E. Lebarbier, C. Levy-Leduc, S. Robin

Maintainer: Souhil Chakar <[e-mail]>

References

S. Chakar, E. Lebarbier, C. Levy-Leduc, S. Robin. A robust approach to multiple change-point estimation in an AR(1) process, arXiv:1403.1958.

Examples

library(AR1seg)
data(y)
res=AR1seg_func(y,Kmax=15,rho=TRUE)
a=c(1,res$PPSelectedBreaks[1:(res$PPselected-1)]+1)
b=res$PPSelectedBreaks[1:(res$PPselected)]
Bounds=cbind(a,b)
mu.fit=rep(res$PPmean,Bounds[,2]-Bounds[,1]+1)
plot(y)
lines(mu.fit,col="red")

Example output

Loading required package: Segmentor3IsBack
Segmentor3IsBack v2.0 Loaded 

AR1seg documentation built on May 2, 2019, 9:51 a.m.